Spinoff: What will happen when process improvements cease?

[superguy]

If the end is really around 30nm?

How, or will, improvements continue?

One thing is, I think the final node will be "perfected" for some time after it is reached. Manufactoring will keep getting better on it. But, this will eventually end, or will it?

You can guess a rough transistor count. They seem to double per node, we're currently at about 400m, so 800 at 65nm, 1.6b at 45 nm, and ~3b at 32nm. Then add increasing efficency perhaps getting us to 5 b.

So we might end up around 10-15 times more powerful than today. Then, that's it! For the most part. Or is it?

 

[bloodbob]

Ehh why is the shrink the end of everything. Lets think about it if you can fit 1 million transistors into 1 mm^2 then in 1 mm^3 you can fit a billion ( you'll have to reduce the clock speed of course ). Changing materials will also get us atleast a 10X increase in clock speeds.

 

[superguy]

Because it's how transistor counts increase?

But if it isn't tell me why.

As for some more thoughts I think improvements can continue past the final node in these ways:

Increased manufactoring effeciency at that node

A higher price on the top end GPU (perhaps $1000?)

Architecture improvements.

It will be weird though. And how will consoles continue to get more powerful?

 

[superguy]

1 mm^3 you can fit a billion

I'm not sure what you mean here..is this possible? A cube shaped chip?

 

[_xxx_]

Give me my crystal ball back and I'll tell you...

 

[bloodbob]

I'm not sure what you mean here..is this possible? A cube shaped chip?

People have made 3d chips it will require completely new fabrication technology but it will happen and it will be most benefical to things like GPUs that can be massively parralled.

 

[superguy]

It would seem to me you would make one by piling a bunch of layers on..seperated by some nonconductive material.

But that doesn't seem to solve you're fundamental problem of cost..it's just stacking current style thin chips on top of each other.

Unless there is some other way..

 

[booomups]

outsourcing... hehe.. no really.


there are already new technologys in development,
i heard something about light being used instead of a current on metal.

and this generation we will see if developers can use several cores in a good way.


but compared to the graphic chips, the slow development of the cpus really worries me.

 

[_xxx_]

there are already new technologys in development,
i heard something about light being used instead of a current on metal.

That's still far away from being deployable. Maybe in 10 years or so.

 

[pc999]

Meybe this kind of tech

http://www.beyond3d.com/forum/showthread.php?t=10218

http://www.beyond3d.com/forum/showthread.php?t=7458

http://www.beyond3d.com/forum/showthread.php?t=7447

It seems very interesting.